[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81702-en":3,"doc-seo-81702-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81702,3848291630094,"Emma Wilson","https://eur-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45",8,"Research & Report","BaRA Budget-constrained and Reliable Web Data Collection Agent","BaRA (Budget-constrained and Reliable Agent) addresses live web data collection where success depends on discovering site-internal pages and retrieving multimodal artifacts within a fixed interaction budget. The framework performs BFS-based link discovery with liveness verification to suppress hallucinated or dead links, then validates extracted text, image, and video using rule-based provenance plus accessibility checks. A history-based self-reflection module recovers from failures and incomplete outputs. Experiments on synthetic and real websites show improved valid-link discovery and download-valid multimodal extraction.","BaRA: Budget-constrained and Reliable Web Data Collection Agent  \nSoojeong Lee * 1 Joseph Lee * 1 Yongseong Cho 2 Sunjae Kim 3 Youngwoo Moon 3 Kyungwoo Song 1  \narXiv :2607 .00007v2 [ cs .IR] 2 Jul 2026  \nAbstract  \nLarge language model (LLM)-based web agents automate web navigation and data collection.  \nHowever, live web data collection demands capabilities beyond task completion: agents must discover site-internal pages and retrieve text, image, and video artifacts in an accessible form within a fixed interaction budget. We formulate this setting as budget-constrained, site-level multimodal web data collection and propose Budgetconstrained and Reliable Agent (BaRA) . BaRA performs breadth-first search (BFS)-based link discovery with liveness verification to filter hallucinated and dead links, then validates extracted multimodal artifacts using rule-based provenance and accessibility checks. A history-based selfreflection module recovers from execution failures and incomplete outputs. On controlled synthetic and real-world websites, BaRA consistently improves valid-link discovery and download-valid multimodal extraction over existing agents. Our code is available at [https://github.com/](https://github.com/)[ ](https://github.com/)MLAI-Yonsei/BaRA-Agent.  \n1. Introduction  \nWeb data collection remains a practical bottleneck for natural language processing (NLP) and multimodal pipelines, including dataset construction, information extraction, and retrieval-augmented generation. LLM-based web agents make interaction with live websites more flexible, but existing web-agent benchmarks mainly evaluate instructionconditioned task completion: whether an agent reaches a target state for a given task. They do not directly target siteinternal page coverage or the validity of collected artifacts.  \n*Equal contribution 1 Department of Statistics and Data Science, Yonsei University, Seoul, Republic of Korea 2 Electronics and Telecommunications Research Institute (ETRI), Daejeon, Republic of Korea 3 HUSTLERS Corp., Seoul, Republic of Korea . Correspondence to: Kyungwoo Song \u003C[kyungwoo.song@yonsei.ac.kr](kyungwoo.song@yonsei.ac.kr) >.  \nProceedings of the Workshop on Planning in the Era of LLMs (LM4Plan) at International Conference on Machine Learning, 2026. Copyright 2026 by the author(s) .  \nIn live web data collection, robust crawling and extraction remain fragile, as agents may miss relevant pages, return incomplete multimodal outputs, or produce media URLs that are hallucinated or not downloadable.  \nWe study a distinct setting: budget-constrained, site-level multimodal web data collection. Given a seed URL, the goal is to discover site-internal pages within a fixed interaction budget and return provenance-grounded text, image, and video artifacts in an accessible form. Unlike taskcompletion benchmarks, this setting treats the validity of discovered site-internal links and extracted artifacts as primary evaluation targets. Although prior benchmarks often impose step or time limits, they typically do not measure whether limited interactions are converted into verified multimodal artifacts. To address this gap, we propose BaRA, a framework for reliable live web data collection.  \nContributions. Our contributions are threefold:  \n• Problem formulation. We formulate budgetconstrained, site-level multimodal web data collection as a distinct evaluation setting for web agents.  \n• Method. We propose BaRA, which orchestrates BFSbased link discovery, liveness verification, reflectionguided recovery, and rule-based artifact verification under a fixed interaction budget.  \n• Benchmarks and evaluation. We introduce two benchmarks: a 50-site synthetic benchmark with known structure and a 50-site real-world benchmark constructed from Tranco (Pochat et al., 2018) longtail domains and Hacker News (Algolia, n.d.) under automatic filtering and diversity constraints. Experiments on both benchmarks show that BaRA improves valid-link discovery and downloa","cbCaigDkHzPszrkZ","https://ap.wps.com/l/cbCaigDkHzPszrkZ","pdf",1180235,3,1,20,"English","en",105,"# Introduction\n# Related Work\n# Method\n## Budget-constrained BFS link discovery\n## Liveness verification and artifact validation\n## History-based self-reflection","[{\"question\":\"What problem does BaRA target in live web data collection?\",\"answer\":\"BaRA targets budget-constrained, site-level multimodal web data collection: discovering site-internal pages and retrieving provenance-grounded text, image, and video artifacts in an accessible form within a fixed interaction budget.\"},{\"question\":\"How does BaRA improve link reliability during crawling?\",\"answer\":\"BaRA uses breadth-first search (BFS) for link discovery combined with liveness verification, filtering out hallucinated and dead links before extraction.\"},{\"question\":\"How does BaRA handle extraction failures or incomplete outputs?\",\"answer\":\"BaRA includes a history-based self-reflection module that rewrites prompts for subsequent attempts, helping recover from execution failures and missing or incomplete results.\"}]",1784175506,50,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"bara-budget-constrained-and-reliable-web-data-collection-agent","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/bara-budget-constrained-and-reliable-web-data-collection-agent/81702/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does BaRA target in live web data collection?","Question",{"text":75,"@type":76},"BaRA targets budget-constrained, site-level multimodal web data collection: discovering site-internal pages and retrieving provenance-grounded text, image, and video artifacts in an accessible form within a fixed interaction budget.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does BaRA improve link reliability during crawling?",{"text":80,"@type":76},"BaRA uses breadth-first search (BFS) for link discovery combined with liveness verification, filtering out hallucinated and dead links before extraction.",{"name":82,"@type":73,"acceptedAnswer":83},"How does BaRA handle extraction failures or incomplete outputs?",{"text":84,"@type":76},"BaRA includes a history-based self-reflection module that rewrites prompts for subsequent attempts, helping recover from execution failures and missing or incomplete results.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,126,129,133],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":29,"slug":113},6,"Technology","technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":22,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":127,"show_sort_weight":22,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":46,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":46,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]